metadata

license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - opus100
metrics:
  - bleu
model-index:
  - name: opus-mt-ar-en-finetuned-ar-to-en
    results:
      - task:
          name: Sequence-to-sequence Language Modeling
          type: text2text-generation
        dataset:
          name: opus100
          type: opus100
          args: ar-en
        metrics:
          - name: Bleu
            type: bleu
            value: 46.8089

opus-mt-ar-en-finetuned-ar-to-en

This model is a fine-tuned version of Helsinki-NLP/opus-mt-ar-en on the opus100 dataset. It achieves the following results on the evaluation set:

Loss: 1.0713
Bleu: 46.8089
Gen Len: 14.1755

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-06
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 16
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
No log	1.0	312	1.2132	43.7663	14.4193
1.3072	2.0	624	1.1869	44.1712	14.4054
1.3072	3.0	936	1.1675	44.5448	14.2182
1.2535	4.0	1248	1.1510	44.8762	14.2004
1.2309	5.0	1560	1.1375	45.2067	14.1375
1.2309	6.0	1872	1.1251	45.4479	14.1887
1.21	7.0	2184	1.1145	45.7117	14.2103
1.21	8.0	2496	1.1051	45.951	14.1665
1.1896	9.0	2808	1.0968	46.1647	14.178
1.1837	10.0	3120	1.0899	46.342	14.1819
1.1837	11.0	3432	1.0842	46.4735	14.1672
1.1589	12.0	3744	1.0795	46.561	14.1729
1.1523	13.0	4056	1.0759	46.6884	14.1706
1.1523	14.0	4368	1.0733	46.7542	14.1735
1.1524	15.0	4680	1.0718	46.7835	14.1712
1.1524	16.0	4992	1.0713	46.8089	14.1755

Framework versions

Transformers 4.19.2
Pytorch 1.7.1+cu110
Datasets 2.2.2
Tokenizers 0.12.1